Associate Machine Learning Engineer at Kargo in New York, NY

Mission 

Build and maintain the ML training, inference, and serving pipelines that power Kargo's auction optimization, outcome prediction, and content recommendation systems. This role takes models from data science prototypes into production environments where they directly influence bid pricing, pacing, CTR prediction, and audience targeting at programmatic scale. Success means reliable, observable, and continuously improving ML systems that move revenue metrics.

This is a hybrid role requiring onsite presence 4 days per week.

Outcomes — What Success Looks Like in 6–12 Months

• Production pipelines shipped: Independently own the deployment of at least 2-3 ML models into production (batch or real-time), including training pipelines, inference endpoints, and monitoring — with documented latency, throughput, and accuracy SLAs met.
• Inference reliability improved: Reduce model serving incidents or latency regressions on owned systems by a measurable margin (e.g., p99 latency, error rate, or successful retraining cadence) compared to baseline at start of tenure.
• MLOps infrastructure contributions: Deliver meaningful improvements to shared ML infrastructure — CI/CD for ML workflows, feature pipelines, model registry, or monitoring tooling — that other engineers and data scientists adopt.
• Cross-functional model launches: Partner with Data Science and Product to productionize at least one revenue-impacting model (auction optimization, CTR/viewability prediction, or recommendation) with clear before/after business metrics.
• Operational ownership: Establish monitoring, alerting, and retraining workflows for production models you own, with documented runbooks and minimal escalations to senior engineers for routine issues.

Skills — Core Technical Capabilities

Required:

• 2-4 years building and deploying ML systems in production; strong Python (C++ a plus)
• Hands-on experience with ML frameworks (PyTorch, TensorFlow, or similar) across the full lifecycle: training, evaluation, deployment, monitoring
• Working knowledge of AWS (SageMaker, EC2, Lambda) or comparable cloud ML services
• Proficiency with large-scale data tooling — SQL, Spark, and Snowflake or equivalent warehouse
• Experience implementing CI/CD pipelines for ML workflows and model serving infrastructure (logging, versioning, lifecycle management)
• Solid grasp of feature engineering, data processing, and debugging production model performance issues

Preferred:

• Experience with model serving frameworks (Triton, FastAPI), containerization (Docker, Kubernetes), or IaC tooling
• Exposure to AdTech, real-time bidding, LLMs/embeddings, or other high-scale distributed systems

Competencies — Behaviors We Like to See

Production Mindset

• Thinks beyond model accuracy to latency, cost, reliability, and observability from day one
• Writes code and pipelines that other engineers can run, debug, and extend without hand-holding

Cross-Functional Translator

• Bridges data science intent and engineering reality — can push back on a model design that won't scale, and explain why in terms a PM understands
• Comfortable pairing with data scientists to harden prototypes and with platform engineers to integrate into the broader stack

Bias for Shipping

• Defaults to small, instrumented launches over large, unmonitored ones; iterates on production signal
• Unblocks themselves on ambiguous problems and asks for help with a clear hypothesis, not a blank slate

Operational Ownership

• Treats models as living systems — sets up monitoring, owns the on-call when something drifts, and follows through on retraining
• Documents what they build so the next person (or the next quarter's Lanissa-era hire) can pick it up cleanly